Robust State Estimation And Monitoring Based On The Characteristics Of Noise Distribution

With the rapid development of modern industry,the scale and complexity of control systems are increasing.Therefore,the requirements for system reliability and safety are increasing.The key and foundation of implementing intelligent control and monitoring is the accuracy of system state.The existing state estimation and monitoring methods are basically constructed under the assumption of Gaussian noise and the noise parameters are known.Due to the influence of the system itself and the environment,the industrial process data often has outliers.Therefore,traditional state estimation methods,such as Kalman filter,often reduce the accuracy of estimation,thereby reducing subsequent process monitoring performance.Considering the different noise distributions and unknown noise parameters in stochastic systems,the state estimation and monitoring methods are proposed in this thesis.The major work of the thesis are as follows.1.Considering the state estimation problem with asymmetric measurement noise and outliers for linear discrete-time system.The skew t distribution is used to characterize the noise properties,and a robust filter algorithm is proposed.By using the variational Bayesian technique,the system state and the noise statistics are estimated simultaneously.The proposed filter can adaptively learn the statistics of measurement noise by a heuristic model,which can be used to track the time-varying measurement noise.A numerical simulation,as well as an experiment on hybrid tank system,is conducted to demonstrate the performance.2.Considering the outliers exist in practice,the t filter that can handle the outliers is extended to the state estimation of the distributed system.A distributed t filtering algorithm is proposed by assuming that the process noise and measurement noise are both t-distributed noise.It is shown that the proposed algorithm provides the same accuracy as the centralized t filtering with no performance loss.Furthermore,the distributed t filtering with feedback is developed,which is in accordance with centralized filtering,and the local error covariance is reduced as expected.Two simulation examples illustrate the effectiveness of the algorithm.3.Considering the sensor fault detection and diagnosis problems for linear systems in the presence of outliers.The t distribution with unknown scale matrix and degrees of freedom(dof)parameter is used to describe the measurement noise.By using the variational Bayesian inference,the states,the scale matrix,and the dof parameter are estimated simultaneously.Since the noise distribution is no longer the Gaussian,a modified residual evaluation is proposed to detect the fault.After that,the cause of fault can be determined by observing the changes on measurement noise covariance.Numerical examples are conducted to demonstrate that the proposed method can provide more reliable results when measurements contain outliers.4.The state estimation and fault estimation problems are considered for linear systems when the statistics of measurement noise are known or unknown.The dynamics of sensor fault are described as a stochastic process.In the probabilistic framework,the potential sensor fault,as well as the system states,is estimated simultaneously by performing the variational Bayesian inference.Different from the existing approaches,the proposed method does not need an accurate description of the fault.Even the statistics of measurement noise are unknown,the proposed method obtains a better fault estimation accuracy.The efficiency and superiority of the proposed method are demonstrated through numerical simulations and experimental tests performed on a hybrid water tank system.